System and Methods for Constructing Personalized Context-Sensitive Portal Pages or Views by Analyzing Patterns of Users' Information Access Activities

1. A computer-readable medium having computer-executable instructions stored thereon to execute a data access model, comprising: a log that collects past web data access patterns; a classifier component to perform automatic topic classification of the past web data access patterns; a context component to record a context relating to the past web data access patterns; and a predictive component to determine future web data access patterns of a user, based at least in part on the past web data access patterns, the topic classification of the past web data access patterns, and the recorded context, and based at least in part on a level of interest I(p) of the user in a page p from among the past web data access patterns, the interest I(p) in the page p being computed as a sum of a first weighted term L(p) indicating a number of links followed from the page p and a second weighted term D(p) indicating time spent in sessions starting from the page p, the first weighted term weighted by a Constant 1 and the second weighted term being weighted b a Constant 2 , the constants selected to equate an average of two links followed from the page p with an average session time length.

2. The model of claim 1 , the predictive component determines a probability of information value to a user given evidence of the user's interest in at least one of a potential site and topic.

4. The model of claim 3 , the evidence including at least one of web pages visited, how much time users spend observing the web pages, how many links the user followed from the web pages, higher-level topic of the web pages viewed, interactivity with the web pages and a consideration of navigation efforts of a user.

5. A computer-implemented method for building a montage, comprising: storing, in a memory, instructions for performing the computer-implemented method for building a montage; executing the instructions on a processor; according to the instructions being executed: logging past data access; logging context information in accordance with the past data access; classifying topic information; building a model to predict future data access of a user, the model based at least in part on the logged past data access, the logged context information, and the classified topic information, and based in part on interest of the user in a page from among the past data access, the interest in the page being computed as I(p)=L(p)*Constant 1 +D(p)*Constant 2 , where I(p) is interest in a page p, L(p) is links followed from page p, D(p) is an average number of seconds spent in sessions starting with page p, and constants Constant 1 and Constant 2 are selected to balance a value based on links followed with a value based on session time length; mining at least one of the past data access and the context information; and selecting data based on the context information.

6. A computer-readable medium having computer-executable instructions stored thereon to execute a data access model, comprising: a log that collects a plurality of past web data access patterns for a computer user, wherein each past web data access pattern comprises a web page accessed by the user on a computer and a context associated with each web page; a classifier component to perform automatic topic classification of the web page accessed by the user a context component to record the context associated with each web site; and a predictive component to determine web future data access patterns based at least in part on the past web data access patterns, the topic classification of the web page accessed by the user, and the recorded context, and based at least in part on interest of the user in a page from among the past web data access patterns, the interest in the page being computed as I(p)=L(p)*Constant 1 +D(p)*Constant 2 , where I(p) is interest in a page p, L(p) is links followed from page p, D(p) is an average number of seconds spent in sessions starting with page p, and constants Constant 1 and Constant 2 are selected to equate an average of links followed from page p with an average session time length of approximately 30 seconds.

7. The model of claim 6 , the predictive component determines a probability of information value to a user given evidence of the user's interest in at least one of a potential site and topic.

9. The model of claim 8 , the evidence includes a web page visited.

10. The model of claim 8 , the evidence includes how much time the user spends observing the web pages.

11. The model of claim 8 , the evidence includes how many links the user followed from the web page.

12. The model of claim 8 , the evidence includes a measure of interactivity with the web page.

13. The model of claim 8 , the evidence includes a consideration of navigation efforts of a user.

14. The model of claim 6 , further comprising the classifier employing a Naive Bayes algorithm.

15. The model of claim 6 , further comprising the classifier employing a Bayes Net algorithm.

16. The model of claim 6 , further comprising the classifier employing a similarity-based algorithm.

17. The model of claim 6 , further comprising the classifier employing a vector-based learning algorithm.